Kaiser":30qd2tid said:
Why are you entering dice into a stat that is definitely not random? Are the chances of a browns QB 1/32 just because there are 32 teams? No, there are reasons.
The reason why people are using the 32-sided dice example here is because it represents an unbiased "outcome" generator. The concept is a valid one to consider if we assume there is no bias against any team. Its the penalties that you would expect if for example, on every play, refs rolled say a six sided die: 1-4 = no penalty, 5 = penalty against Team A, 6 = penalty vs team B.
In statistics it might represent the "null hypothesis". ie. it is what you expect if there is no bias at all. In this case, if there was no bias in how the refs penalise teams throughout the year, you would expect your observations over the years to match what you would expect if the penalties were in fact determined by a random die roll. Of course we know all teams/players/coaches don't all play the same way so differences are EXPECTED between the penalties against each team.
The actual statistic of interest here however is not however as simple/straight forward one to "see"/understand amongst the typical NFL stats.
Breaking it down, you would have to do it like this: At the end of each year, for each team that played Seattle, take the total penalties they incurred for the full year and divide it by the number of games they play to get a penalties/game average. Now subtract this yearly penalties/game average from the actual gameday penalties they incurred in each game when they played Seattle. If teams played Seahawks multiple times, take the average of the values calculated across those games.
You will now have a number for each team that played Seattle in a season that represents how penalised they were when playing Seattle as compared to their penalties per game average across the whole season. A POSITIVE number would mean they had been penalised MORE than their yearly per game penalty average when playing Seattle; if it is NEGATIVE, then they had been penalised LESS when playing Seattle compared to their yearly per game penalty average,
If you now take these numbers and average them all up and assume there is no bias in the way teams are penalised, the EXPECTED number for this would be 0. If teams are getting penalised LESS than they normally do when playing Seattle (which is what is being said), then the EXPECTED number would be NEGATIVE.
Note: the concept of EXPECTED number in statistics is a useful one to consider. It's what you expect based on what you assume, typically if you repeated "the experiment" an infinite amount of times.
You would then do this for every other team and compare to Seattle.
If we find over the last few years that this value is NEGATIVE for Seattle, then it is clear that teams ARE getting penalised less against Seattle compared to their penalties/game across the whole season.
Further if we now rank this Seattle value against the values calculated for each other team in the NFL across a few years and find Seattle is NOT appearing to vary randomly (eg. is typically low or bottom ranking), then it further highlights a potential bias in outcome.